Just published search_ash 0.5.0 (with search_core 0.4.0) on Hex.
What’s new: synonyms
You can now declare a synonym dictionary per language:
elixir
synonyms %{en: %{"cat" => ["kitten"]}}
Typing cat now builds a tsquery like (cat | kitten), matching documents whose search_text contains either. A few things worth knowing:
- It’s dictionary-based, not semantic — the library doesn’t “know”
catandkittenare related, it applies exactly the mapping you give it. - It’s one-directional:
"cat" => ["kitten"]means typingcatalso findskitten, not the reverse — add both directions if you want symmetry. - Stemming still applies on both sides, so plurals/inflections are handled automatically.
- Synonyms only expand the full-text match + ranking (
search_text/ts_rank) — they deliberately don’t affect the global-search label ranking tiers or fuzzy/typo matching, so an exact title match still ranks first.
This turns out to be pretty handy for domain acronyms too ("bl" => ["bon de livraison"], etc.) — not just lexical synonyms.
What’s fixed: a memory leak
Also fixed a subtle memory retention issue related to how Elixir handles large binaries — a small referenced binary was keeping a much larger one alive longer than necessary. Fixed by forcing a copy at the right point in the pipeline instead of holding a sub-binary reference.
Feedback welcome, as always — especially on the synonym API shape if anyone has strong opinions on bidirectional-by-default vs explicit.
repo: https://github.com/nseaSeb/search_ash
